Skip to main content
eLife logoLink to eLife
. 2021 Jul 12;10:e64683. doi: 10.7554/eLife.64683

Ct threshold values, a proxy for viral load in community SARS-CoV-2 cases, demonstrate wide variation across populations and over time

A Sarah Walker 1,2,3,4,, Emma Pritchard 1,2, Thomas House 5,6, Julie V Robotham 2,7, Paul J Birrell 7,8, Iain Bell 9, John I Bell 10, John N Newton 11, Jeremy Farrar 12, Ian Diamond 9, Ruth Studley 9, Jodie Hay 13,14, Karina-Doris Vihta 1,2, Timothy EA Peto 1,2,3,15, Nicole Stoesser 1,2,3,15,, Philippa C Matthews 1,15,, David W Eyre 1,2,14,16,, Koen B Pouwels 1,2,17; COVID-19 Infection Survey team
Editors: M Dawn Teare18, Miles P Davenport19
PMCID: PMC8282332  PMID: 34250907

Abstract

Background:

Information on SARS-CoV-2 in representative community surveillance is limited, particularly cycle threshold (Ct) values (a proxy for viral load).

Methods:

We included all positive nose and throat swabs 26 April 2020 to 13 March 2021 from the UK’s national COVID-19 Infection Survey, tested by RT-PCR for the N, S, and ORF1ab genes. We investigated predictors of median Ct value using quantile regression.

Results:

Of 3,312,159 nose and throat swabs, 27,902 (0.83%) were RT-PCR-positive, 10,317 (37%), 11,012 (40%), and 6550 (23%) for 3, 2, or 1 of the N, S, and ORF1ab genes, respectively, with median Ct = 29.2 (~215 copies/ml; IQR Ct = 21.9–32.8, 14–56,400 copies/ml). Independent predictors of lower Cts (i.e. higher viral load) included self-reported symptoms and more genes detected, with at most small effects of sex, ethnicity, and age. Single-gene positives almost invariably had Ct > 30, but Cts varied widely in triple-gene positives, including without symptoms. Population-level Cts changed over time, with declining Ct preceding increasing SARS-CoV-2 positivity. Of 6189 participants with IgG S-antibody tests post-first RT-PCR-positive, 4808 (78%) were ever antibody-positive; Cts were significantly higher in those remaining antibody negative.

Conclusions:

Marked variation in community SARS-CoV-2 Ct values suggests that they could be a useful epidemiological early-warning indicator.

Funding:

Department of Health and Social Care, National Institutes of Health Research, Huo Family Foundation, Medical Research Council UK; Wellcome Trust.

Research organism: Other

Introduction

After initial reductions in SARS-CoV-2 cases in mid-2020, following release of large-scale lockdowns (Flaxman et al., 2020), infection rates have undergone waves of resurgence and suppression in many countries worldwide. Proposed control strategies include new local or national lockdowns of varying intensity and mass testing, but these have major economic and practical limitations. In particular, mass testing of large numbers without symptoms (Yokota et al., 2020), and hence low pre-test probability of positivity, can mean most positives are false-positives depending on test specificity. For example, with 0.1% true prevalence, testing 100,000 individuals with a 99.9% specific test with perfect sensitivity gives 100 true-positives, but also 100 false-positives (positive predictive value [PPV] 50%), whereas specificity of 99.5% increases false-positives to 500 (PPV = 17%), and of 99.0% to 999 (PPV = 9%), with even lower PPV with imperfect sensitivity (Adams et al., 2020).

Mathematical models are powerful tools for evaluating the potential effectiveness of different control strategies, but rely on population-level estimates of infectivity and other parameters. However, there are few unbiased community-based surveillance studies, including individuals both with and without symptoms. Estimates of asymptomatic infection rates vary, being 17–41% overall in recent reviews (Buitrago-Garcia et al., 2020; Byambasuren et al., 2020), but these included many studies of contacts of confirmed cases. Higher prevalence of asymptomatic infection has been reported in screening of defined populations (30% [Buitrago-Garcia et al., 2020]) and community surveillance (e.g. 42% Lavezzo et al., 2020, 72% Riley and Ainslie, 2020a). Studies have generally indicated lower rates of transmission from asymptomatic infection (Buitrago-Garcia et al., 2020; Byambasuren et al., 2020), this may be a proxy for SARS-CoV-2 viral load as a key determinant of transmission. Finally, most studies rely on ‘average’ estimates of the asymptomatic infection percentage, independent of characteristics and viral load, and have not quantified temporal variation in these key parameters for mathematical models across the community.

Here we therefore characterise variation in SARS-CoV-2-positive tests in the first 11 months of the UK’s national COVID-19 Infection Survey. In brief (details in Materials and methods), the survey randomly selects private households to provide a representative UK sample, recruiting all consenting individuals aged 2 years or older currently resident in each household to provide information on demographics, symptoms, contacts and relevant behaviours and self-taken nose and throat swabs for RT-PCR testing (Pouwels et al., 2021). A randomly selected subset is approached for additional consent to provide blood samples for IgG S-antibody testing if aged 16 years or older. At the first visit, participants can provide additional consent for longitudinal follow-up (visits every week for the next month, then monthly for 12 months from enrolment). We estimate predictors of RT-PCR cycle threshold (Ct) values (as a proxy for viral load), propose a classification for the strength of evidence supporting positive RT-PCR test results in the community, and demonstrate how this has changed over time. We also provide a preliminary assessment of seroconversion rates for community cases.

Materials and methods

This study included all positive SARS-CoV-2 RT-PCR results between 26 April 2020 and 13 March 2021 from nose and throat swabs taken from participants in the Office for National Statistics (ONS) CIS (ISRCTN21086382). The survey randomly selects private households on a continuous basis from address lists and from previous surveys to provide a representative UK sample (Supplementary file 1). If anyone aged 2 years or older currently resident in an invited household agreed verbally to participate, a study worker visited the household to take written informed consent, which was obtained from parents/carers for those 2–15 years; those aged 10–15 years provided written assent. The study protocol is available at https://www.ndm.ox.ac.uk/covid-19/covid-19-infection-survey/protocol-and-information-sheets. Recruitment started 26 April 2020 in England, 29 June 2020 in Wales, 29 July 2020 in Northern Ireland, and 21 September 2020 in Scotland.

Individuals were asked about demographics, symptoms, contacts, and relevant behaviours (https://www.ndm.ox.ac.uk/covid-19/covid-19-infection-survey/case-record-forms). To reduce transmission risks, self-taken nose and throat swabs were obtained following study worker instructions. Parents/carers took swabs from children under 12 years. At the first visit, participants were asked for (optional) consent for follow-up visits every week for the next month, then monthly for 12 months from enrolment. In a random 10–20% households, those 16 years or older were invited to provide venous blood monthly for assays of anti-trimeric spike protein IgG using an immunoassay developed by the University of Oxford (National SARS-CoV-2 Serology Assay Evaluation Group, 2020). All participants in households where anyone tested positive on a swab were also invited to provide blood monthly. Venous blood was not taken at any visit where any person in the household had classic COVID-19 symptoms (fever, cough, or anosmia/ageusia). The study received ethical approval from the South Central Berkshire B Research Ethics Committee (20/SC/0195).

Swabs and blood samples were collected by study workers at household visits and couriered overnight to testing laboratories at ambient temperatures. They were analysed at the UK’s national Lighthouse Laboratories at Milton Keynes (National Biocentre) (from 26 April 2020 to 11 February 2021) and Glasgow (from 16 August 2020) using identical methodology, with swabs from specific regions sent consistently to one laboratory. RT-PCR for three SARS-CoV-2 genes (N protein, S protein, and ORF1ab) used the Thermo Fisher TaqPath RT-PCR COVID-19 Kit, analysed using UgenTec Fast Finder 3.300.5 (TaqMan 2019-nCoV Assay Kit V2 UK NHS ABI 7500 v2.1). The Assay Plugin contains an Assay-specific algorithm and decision mechanism that allows conversion of the qualitative amplification Assay PCR raw data from the ABI 7500 Fast into test results with minimal manual intervention. Samples are called positive in the presence of at least single N gene and/or ORF1ab but may be accompanied with S gene (one, two, or three gene positives). There is no specific Ct threshold for determining positivity. S gene is not considered a reliable single-gene positive (as of mid-May 2020). Blood was analysed at the University of Oxford. Antibody titres were considered positive above 8 million units (National SARS-CoV-2 Serology Assay Evaluation Group, 2020) on the original fluorometric version of the assay and 42 units on the colorimetric version of the assay (used from 1 March 2021).

Twelve specific symptoms were elicited at each visit (cough, fever, myalgia, fatigue, sore throat, shortness of breath, headache, nausea, abdominal pain, diarrhoea, loss of taste, loss of smell), as was whether participants thought they had (unspecified) symptoms compatible with COVID-19. From 26 April through 22 July 2020, questions referred to current symptoms, and from 23 July 2020 to the preceding 7 days. Any positive response to any symptom question at the swab-positive visit defined the case as symptomatic ‘at’ the test; we also separately defined any positive response at the swab-positive visit or visits either side (regardless of time between visits) as symptomatic ‘around’ the test.

To investigate the potential increasing contribution of false-positives as population prevalence declines, from 2 August 2020 we arbitrarily classified in real-time each positive as:

  • ‘Higher’ evidence: two or three genes detected (irrespective of Ct).

  • ‘Moderate’ evidence: single-gene detected and (1) Ct below the 97.5th percentile of ‘higher’ evidence positives (<34; supporting this threshold, whole genome sequences had been obtained from three single-gene positives with Ct 30.8–33.1 by 2 August) or (2) higher pre-test probability of infection, defined as any symptoms at/around the test or reporting working in a patient-facing healthcare or care/residential home.

  • ‘Lower’ evidence: all other positives; by definition single-gene detected at Ct ≥ 34 in individuals not reporting symptoms/working in relevant roles.

As the Ct distribution was skewed to the left, we assessed independent predictors using median (quantile) regression. Results were broadly similar using random effects model for mean Ct with a random effect per household. We used five knot natural cubic splines (knots at the 10th/25th/50th/75th/90th percentiles of observed unique values) to assess non-linearity in the effect of calendar time, age, and deprivation (index of multiple deprivation rank). Multivariable models for Ct values were constructed by first choosing the more strongly univariably predictive factor from the collinear variables (symptoms at/around the test, number of genes detected/supporting evidence for each positive) and then using backwards elimination on the remaining variables. Deprivation was assessed using the index of multiple deprivation (IMD) in England, a score based on lower layer super output areas with average population of 1500 people and incorporating seven domains to produce an overall relative measure of deprivation (income, employment, education, skills and training, health and disability, crime, barriers to housing services and living environment) (https://www.gov.uk/government/statistics/english-indices-of-deprivation-2019) and equivalent scores in the other three countries comprising the UK. Each country’s scores were converted to a within-country percentile. All analyses were conducted in Stata 16.1.

Results

Number and percentage of positive swabs

From 26 April 2020 to 13 March 2021, 440,479 participants from 217,887 households in the COVID-19 Infection Survey had one or more RT-PCR results from nose and throat swabs (median eight results per participant [IQR 6–9, range 1–19]). Participants were recruited between April 2020 and March 2021 (Supplementary file 1). Of 3,312,159 RT-PCR test results, 27,902 (0.84%, 95% CI 0.83–0.85%) were positive, in 21,831 individuals from 16,214 households. Two thousand nine hundred and sixty-six (14%) of these individuals were positive at their first test in the study and 18,865 (86%) subsequently, after median five negative tests (IQR 3–6, range 1–14).

Viral characteristics

Overall, 10,317 (37%), 11,012 (40%), and 6550 (23%) swabs were positive for three, two, or one of the three SARS-CoV-2 genes (N protein, S protein, and ORF1ab), respectively (Table 1; 23 positives with missing Ct and gene detection excluded from this and all subsequent analysis; samples with only the S-gene detected generally not called positive, see Materials and methods). The majority of two-gene positives (9513 [86%]) were ORF1ab+N positive from 16 November 2020 onwards, reflecting the emergence and expansion of B.1.1.7 (WHO Alpha) in the UK (Walker et al., 2021). B.1.1.7 leads to S-gene target failure (SGTF) and was estimated to account for 88% of SGTF from this time (Public Health England, 2020). Where multiple genes were detected, the Cts were highly correlated (Spearman rho = 0.98, p<0.0001). Taking the per-swab mean Ct across positive genes, the overall median Ct was 29.2 (IQR 21.9–32.8; range 9.2–38.7), reflecting the study’s surveillance design testing individuals in the community at fixed timepoints regardless of symptoms. Based on calibration data (Appendix 1—figure 1), this corresponds to a median viral load of ~215 copies/ml (IQR 14–56,400). Ct varied strongly by number of genes detected (Kruskal–Wallis p=0.0001), but not by their specific pattern after adjusting for number (p=0.08). There is no fixed Ct threshold for determining positivity (see Materials and methods); however, only 38 (0.1%) Ct values > 37 were recorded (five positive on ORF1ab+N).

Table 1. Genes detected in positive swabs.

All positives (N = 27,879) First positive per participant (N = 21,811)
Number of genes detected N (%) Median CT* (IQR) [range] N (%) Median CT* (IQR) [range]
1 6550 (23%) 33.8 (32.9–34.7) [12.7–38.7] 5102 (23%) 33.9 (32.9–34.7) [12.7–38.7]
2 1145 (4%) 32.3 (30.9–33.4) [10.3–37.2] 773 (4%) 32.3 (30.7–33.4) [10.3–37.2]
2: ORF1ab+N 16 Nov 2020 onwards 9867 (35%) 26.4 (19.4–31.1) [9.2–37.8] 8184 (38%) 25.3 (18.6–30.7) [9.2–37.8]
3 10,317 (37%) 25.3 (19.8–29.5) [9.3–36.8] 7752 (36%) 23.9 (18.8–28.8) [9.3–36.8]
Genes detected
N only 4479 (13%) 33.9 (33.0–34.8) [26.1–38.7] 3419 (16%) 34.0 (33.1–34.8) [28.2–38.7]
ORF1ab only 2044 (7%) 33.6 (32.6–34.5) [16.8–38.3] 1656 (8%) 33.7 (32.7–34.6) [16.8–38.3]
S only† 27 (0.1%) 34.9 (33.5–36.1) [12.7–37.3] 27 (0.1%) 34.9 (33.5–36.1) [12.7–37.3]
N+ORF1ab: before 16 Nov 2020 731 (3%) 31.9 (30.3–32.9) [10.3–37.2] 497 (2%) 31.8 (29.7–33.0) [10.3–38.2]
N+ORF1ab: 16 Nov 2020 onwards 9867 (35%) 26.4 (19.4–31.1) [9.2–37.8] 8184 (38%) 23.9 (18.8–28.8) [9.3–36.8]
S+ORF1ab 190 (0.7%) 32.5 (31.2–33.5) [15.1–36.6] 138 (0.6%) 32.4 (31.0–33.6) [15.1–36.6]
N+S 224 (0.8%) 33.4 (32.5–34.2) [25.0–36.8] 138 (0.6%) 33.3 (32.4–34.3) [27.3–36.8]
N+S+ORF1ab 10,317 (37%) 25.3 (19.8–29.5) [9.3–36.8] 7752 (36%) 25.3 (18.6–30.7) [9.2–37.8]

*Taking the mean Ct per positive swab across positive gene targets (Spearman rho = 0.98 for each pair of genes where both positive, p<0.0001).

†17/27 before mid-May only: after this samples positive for the S gene only were not called positive overall by the algorithm and therefore reflect likely recording errors.

Note: excluding 23 positive results without Ct values or genes detected available. Comparing first vs subsequent positives per participant, exact p<0.0001 for both number of genes detected and specific genes detected.

Of note, whilst single-gene positives almost invariably had Ct>30, with or without reported symptoms, triple-gene positives without reported symptoms had widely varying Ct, as did ORF1ab+N positives after 16 November 2020 (SGTF, compatible with B.1.1.7) (Figure 1). Ct values were slightly but significantly lower in other double-gene positives vs single-gene positives, with a small number of low Ct values in ORF1ab+N positives before 16 November 2020 likely reflecting early B.1.1.7 cases. Furthermore, whilst the percentage reporting symptoms increased linearly as Ct values dropped from 35 (~30% reporting symptoms around the positive test) to 28 (~60% reporting symptoms), below 28 the percentages reporting symptoms increased only slightly (to ~70% at Ct=10) (Figure 2).

Figure 1. Distribution of Ct values at each positive test by genes detected and self-reported symptoms.

Figure 1.

Note: Points show the median and boxes the interquartile range. OR=ORF1ab. Positives where only the ORF1ab+N genes were detected are split by whether the swab was taken before or after 16 November 2020, reflecting the expansion of B.1.1.7 (which has S-gene target failure on the assay used in the survey).

Figure 2. Percentage reporting symptoms by Ct value.

Figure 2.

Note: Points show the percentage of positive tests with each rounded Ct value reporting any symptoms or cough, fever, anosmia/ageusia at each test or around each test (see Materials and methods for symptoms collection and definitions). Ct values under 11 and over 36 grouped with 11 and 36, respectively.

Evidence supporting positive results

Combining information on Ct values, symptoms and pre-test probability of being positive, 21,329 (77%), 4741 (17%), and 1809 (6%) positive tests had ‘higher’, ‘moderate’, or ‘lower’ evidence supporting genuine presence of viral RNA (Table 2; definitions in Materials and methods). Even though ‘higher’ evidence was based only on number of genes detected (two or three), ‘higher’ evidence positives were more likely to be symptomatic than ‘moderate’ evidence positives (p<0.0001), but were similarly likely to have occupational risk factors (p=0.48). ‘Higher’ evidence positives were more likely to occur in households with other positives (p<0.0001).

Table 2. Evidence supporting positive test results indicating presence of virus and impact on other factors.

Strength of evidence for true infection
Higher Moderate Lower p (exact)
Number (col %) (N = 27,879) 21,329 (77%) 4741 (17%) 1809 (6%)
Factors determining classification
Number of genes detected
(row %)
3: 10,317 (48%)
2: 11,012 (52%)
1: 4741 (100%) 1: 1809 (100%)
CT, median 26.2 33.4 34.8
CT, n (row %) <34* 21,070 (98.8%) 3613 (76%) 0 (0%)
Symptoms around test, n (row %) 12,466 (58%) 2243 (47%) 0 (0%) <0.0001
(exc lower)
Occupational risk, n (row %) 1322 (6%) 307 (6%) 0 (0%) 0.48
(exc lower)
Other factors
Cough, fever, anosmia, ageusia around test, n (row %) 9345 (44%) 1241 (26%) 0 (0%) <0.0001
(exc lower)
First positive test n (row %) (vs subsequent positive test) 16,709 (78%) 3508 (74%) 1594 (88%) <0.0001
First test in study, n (row %) (vs follow-up i.e. prior negative in study) 2281 (11%) 482 (10%) 199 (11%) 0.49
Any genome sequence obtained, confirming presence of virus‡ 6,621/9,022 (73%) 544/2,315 (24%) 0/836 (0%) <0.0001
Any other household member ever positive$ 11,493/18,494 (62%) 1,513/4,004 (38%) 318/1,525 (21%) <0.0001

*Approximate 97.5th percentile of CT in higher evidence positives through 2 August 2020 when classification first applied.

Reported working in a patient-facing healthcare role/care/residential home.

Any genome sequence obtained out of attempted (other positives not found or not yet attempted).

$Denominator households with two or more study participants.

Note: Classification arbitrarily determined on 2 August 2020 based on the number of genes detected, Ct values and pre-test probability (see Materials and methods).

Predictors of Ct values

In multivariable regression models, Ct values were independently lower (i.e. viral loads higher) with more genes detected (8.2 lower in triple-gene vs single-gene positives [95% CI 7.9–8.5]), if symptoms were reported around the test (2.0 lower [1.8–2.2]), at the first positive identified per participant (2.2 lower than subsequent positives [2.2–2.5]), and if the positive was not the participant’s first test in the study (0.6 lower [0.2–0.9]) (all p<0.0001; Supplementary file 2A; see Materials and methods for details of collection of symptoms). By far the strongest effect was associated with triple-gene positives. Men had slightly lower Ct values than women (0.3 lower [0.1–0.5] p=0.001), and there was marginal evidence of lower values in those reporting non-white ethnicity (0.3 lower [0–0.6] p=0.08). Compared with those not reporting symptoms, Ct values were lower in those reporting cough/fever/anosmia/ageusia (2.5 lower [2.3–2.8]) than other symptoms only (0.9 lower [0.7–1.2]; heterogeneity p<0.0001). Associations were similar for symptoms at the positive test. After adjusting for these factors, there was no evidence of independent effects of age (p=0.33) or deprivation (p=0.67, Supplementary file 2A). Even after adjusting for these factors, Ct values were 1.4 (1.2–1.6) lower in individuals where another household member was positive at any point in the study (p<0.0001; other effects similar).

However, number of genes detected and symptoms are both potential mediators of effects of demographic factors (Appendix 1—figure 2). Excluding these potential mediators (number of genes detected, symptoms), Ct values remained independently lower (i.e. viral loads higher) at the first positive identified per participant, where the positive was not the participant’s first test in the study, and in men, but were also slightly lower with increasing deprivation (p=0.0005; Ct 1.0 lower in the most vs least deprived [95% CI 0.6–1.5]) and in younger adults (p=0.0001; those aged 17–24 1.0 lower [0.3–1.7] than those under 12, and 1.4 lower [0.8–2.0] than those aged 70+) (Supplementary file 2B). Results were similar adjusting for date of the positive test.

Temporal changes in Ct values, evidence, and symptomatic percentages

There were strong effects of calendar time on the distribution of Ct values (Figure 3A,B), the percentages self-reporting symptoms, or cough/fever/anosmia/ageusia (Figure 3C), and strength of evidence supporting each positive result (Figure 3D; all p<0.0001). In particular, Ct values were markedly higher in July–August 2020 when population positivity rates were low, with correspondingly very low percentages with symptoms at/around positive tests, and more ‘lower’ evidence positives. Decreases in Ct values in late August/early September and December 2020 coincided with increases in percentages reporting symptoms and of ‘higher’ evidence positives, and, in England (Figure 3B), with initial rises in official estimates of positivity rates (Office for National Statistics, 2021) after very low rates in July/early August 2020, and with much stronger rises in December 2020 (expansion of B.1.1.7). Ct levels rose, and correspondingly percentages reporting symptoms and of ‘higher’ evidence positives declined, as positivity peaked during November 2020 and January 2021 lockdowns.

Figure 3. Variation over calendar time in the distribution of Ct values in the UK (A) and England (B) together with percentage positivity in England (B), and in self-reported symptoms (C) and evidence supporting positives (D).

Figure 3.

Note: Panel (A) shows the distribution of Ct values each week including all positives across the UK. Panel (B) is restricted to England shown together with the official estimates of positivity as reported by the Office for National Statistics (black line) and periods of national ‘stay-at-home’ restrictions (schools shut in dark grey, schools open in light gray). Panels (C) and (D) show the proportions reporting symptoms and with different levels of evidence supporting the positive test, respectively. Variation in the width of 95% CI reflects the increase in size of the survey from mid August (Supplementary file 1).

However, even within ‘higher’ evidence positives, median Ct varied strongly over time being higher in July/early August 2020 and after November 2020 and January 2021 lockdowns (Figure 4A). ‘Lower’ evidence positives also formed a larger percentage of all tests during July/early August 2020, despite overall positivity rates being very low (e.g. 0.022% in the 3 weeks starting 20 July 2020; Figure 4B). However, interestingly, from September 2020, the percentage of ‘lower’ evidence positives increased proportionately with ‘moderate’ and ‘higher’ evidence positives (Figure 4B). The lowest non-zero observed rate of ‘low evidence’ positives was 0.005% (both in early June and late August), providing an upper bound on the rate of false-positives as defined by identifying virus when none present.

Figure 4. Ct values (A) and percentage positive of all tests (B) by level of evidence and time.

Figure 4.

Note: Panel (A) shows median Ct values according to level of evidence and panel (B) percentage of all swab tests positive according to level of evidence over calendar time. The early part of the study is grouped into 3 week periods due to lower numbers of positives.

Relationship with serostatus

One or more IgG S-antibody results were available for 6540 (30%) participants with positive swabs. Less than 5% of antibody tests taken >30 days before the first positive swab (not necessarily the onset of infection) were positive (Figure 5), rising to 12% in the 30 days before the first swab positive (likely reflecting late detection of infection), 47% in the following 14 days, and then 72–81% thereafter. Overall, of 6189 participants with one or more antibody tests after their first positive swab, 4808 (78%) were ever antibody-positive; with higher rates in those reporting symptoms around their first positive swab (2945/3315 [89%] vs 1863/2874 [65%] of those not reporting symptoms, p<0.0001). Median (IQR) Ct values were also significantly lower in those ever antibody-positive to date (24.9 [18.5–31.0] vs 33.0 [29.9–34.3] in those not antibody-positive, p<0.0001). Results were similar restricting to 1477 (24%) with a negative antibody result within [−120, +21] days of their first positive swab. A small number of participants appeared to have become infected despite antecedent high anti-spike antibody titres, one case in particular which had ‘higher evidence’ positive swab tests separated by four consecutive negative swabs with 65 days between positive swabs.

Figure 5. Percentage of positive antibody tests over time from first positive swab.

Figure 5.

Note: showing the percentage of participants with S-antibody positive or negative tests according to days from their first positive swab, separately for those with and without any antibody results prior to their first positive swab.

Discussion

In this large community surveillance study, we found wide variation in Ct values (a proxy for viral load). Whilst Ct values were independently associated with several factors, including symptoms at/around the test as previously reported (Edwards et al., 2020; Lee et al., 2020), their effects were small compared with population-level variability. Notably both triple-gene positives and S-gene target failures compatible with the Alpha/B.1.1.7 variant without reported symptoms had widely varying Ct, including many with low values (Figure 3A), potentially explaining variation in dispersion (‘k’) and super-spreading events, particularly from those without symptoms but with low Ct/high viral loads (Endo et al., 2020; Rasmussen and Popescu, 2021).

Compared with other single/double positives, Ct values were significantly lower in triple-gene positives and S-gene target failures compatible with the B.1.1.7 variant (after mid-November 2020). However, direct comparisons of viral load with B.1.1.7 vs other variants are not possible within this analysis, given lack of knowledge as to the true underlying variant over the included time period. We found lower Ct in those reporting cough/fever/anosmia/ageusia than other symptoms, and other symptoms vs no symptoms, supporting the importance of the ‘classic’ symptoms for identifying the most infectious cases. Lower Ct values in the first positive per participant likely reflects the natural history of viral load post-infection, and higher Ct values in those positive at their first test in the study over-representation of long-term shedders in this group. Lower Ct values in men and those reporting non-white ethnicity, although small, are consistent with poorer outcomes in these groups. Interestingly, small effects of age and deprivation were mediated by self-reported symptoms and number of positive genes. That is, when adjusting for these latter factors no association was observed with Ct, but without adjustment younger individuals (as shown in Jones et al., 2020 but not Jacot et al., 2020) and those from more deprived areas had slightly lower Ct values, suggesting that these factors may affect the intrinsic level of virus present (Appendix 1—figure 2). The small size of the effects mean they may variably be detected depending on study size and power.

Ct values varied strongly over time, as did symptoms and evidence supporting positives, suggesting changing viral burden in infection cases, with less severe infections during July/early August 2020. This strongly refutes hypotheses that declines in positivity during this period were due to declines in viral fitness. During this time, higher Ct values were also noted in the English point-prevalence surveillance study, REACT (Riley and Ainslie, 2020b), and lower virus levels in Lausanne, Switzerland (Jacot et al., 2020). However, Ct values were higher even in ‘higher’ evidence positives during this period, consistent with shifting viral burden (Figure 4A). Such a shift may also explain the preceding shift towards ‘moderate’ evidence positives and the concurrent higher percentage of ‘lower’ evidence positives, since the less virus present, the less likely it is to be detected on multiple genes. Whilst these findings are consistent with lower viral inoculum during this period (Gandhi et al., 2020), we cannot assess whether this is predominantly due to behaviour (e.g. increased time outdoors, face mask use Gandhi and Rutherford, 2020) or other reasons (e.g. environmental/climatic factors, including relating to transport of swabs for testing). Whilst decreases in Ct values in July/early August 2020 preceded increases in positivity rates in England, later declines in Ct in early December coincided with, rather than preceded, increases in positivity due to B.1.1.7 expansion. This may potentially reflect faster transmission of B.1.1.7 but may also reflect greater sensitivity to changes in Ct distribution when case numbers are small. Subsequent increases in Ct reflected stabilising and then declining positivity in both periods.

We used laboratory, clinical, and demographic evidence to classify our confidence in positive results. Around 70% had two or three genes detected (‘higher’ evidence), providing assurance in overall results, with only 0.1% of Ct values over 37. Whilst Ct values are not directly comparable between studies, REACT has also validated a Ct threshold of 37 for single-gene positives for their test performed in Germany (Riley and Ainslie, 2020b), and in the Public Health England (PHE) Schools study, only samples with Ct<37 were positive on repeat testing of the same swab at PHE laboratories (Ladhani et al., 2020). However, every diagnostic test has false-positives, here defining a false-positive as detection of virus by RT-PCR when no virus is present in a sample, so some of our single-gene ‘lower’, or even ‘moderate’, evidence positives are inevitably false. However, the false-positive rate (as defined) would generally be expected to be approximately constant over time, since it is either random or driven by external factors, although cross-contamination (which should be minimised by good laboratory practice) may theoretically be related to background prevalence/viral load. Variation in the percentage of all tests accounted for by ‘lower’ evidence positives, and in particular the proportionate increases in ‘lower’ evidence positives as ‘higher’ evidence positives increased during September 2020 supports more genuinely lower-level infections occurring during the summer, and an overall false-positive rate for this test of below ~0.005% that is at least 99.995% specificity.

With recent expansion of antigen assays, there has been considerable debate on what ‘positivity’ means, and hence what is a ‘false-positive’ or a ‘false-negative’. First, it is clear that the detection of viral RNA is neither the same as infectiousness, although a strong relationship between Ct values and infection in contacts is observed (Lee et al., 2021), nor a ‘disease’ in its own right. However, surveillance has very distinct goals from clinical testing with its focus on isolation and contact tracing, particularly given the large percentage of asymptomatic infections. It is appropriate for surveillance to focus on detection of viral RNA, given its goal to estimate burden of current/ongoing cases that have occurred in the community. However, it is essential to recognise the difference between the RT-PCR test result (viral RNA has been detected) and the appropriate clinical action, which may legitimately differ depending on Ct value, for example if the infection is likely to have occurred sometime previously, as well as other information (e.g. preceding PCR positivity or serology). RT-PCR assays test for viral RNA presence, and hence it is more relevant to consider limits of detection, rather than ‘false-positives’ per se. Although they were a small minority (6%), one question is whether single-gene positives with high Ct (defined as ≥34 in our study) solely represent long-term shedding of non-transmissible virus (Moraz et al., 2020), with, for example, infectious virus recovered from only 8% (95% CI 3–18%) of samples with Ct>35 in a PHE study (Singanayagam et al., 2020) and studies reporting no growth of virus for Ct thresholds from >24 to >34 or higher (Jefferson et al., 2020). Whilst we have not directly assessed household transmission in this analysis, it was notable that Ct values were significantly lower in positives where anyone else in the same household was ever positive, supporting a role for greater within-household transmission with lower Ct values. Ct values were 0.6 higher in positives that were a participant’s first study test (where long-term shedders would be expected to be overrepresented), but these formed only 14% of the positives.

Our evaluation of serological responses is one of few in the community to our knowledge and highlights that a significant minority (~20%) of RT-PCR-positive cases do not appear to seroconvert, particularly those with higher Ct values and not reporting symptoms. A recent systematic review estimated that 95% of adults with laboratory confirmed SARS-CoV-2 infection developed IgG antibodies (Arkhipova-Jenkins et al., 2021), peaking around 25 days. However, only 23% of included studies were in outpatient settings and 14% included only participants with asymptomatic or mild disease. Our community setting, with higher percentages not reporting symptoms and higher Ct values (both associated with not seroconverting), likely explains our lower overall seroconversion estimate compared with these previous studies. We observed a small number of new swab positives in antibody-positive individuals: unfortunately whole-genome sequence data were not available to confirm potential re-infections. Presumed re-infections have been reported elsewhere (Tomassini et al., 2021), including in individuals without previous functional and/or durable antibody responses (Goldman et al., 2020; To et al., 2020), and may remain relevant to virus transmission, whether they occur with or without symptoms. Our data and others (Lumley et al., 2021) suggest that these may occur in the presence of anti-spike antibodies, which correlate with neutralising antibody titres. These antibody titres are unlikely to have been false-positives, given the context, persistence, and known diagnostic and analytical specificity of the assay (National SARS-CoV-2 Serology Assay Evaluation Group, 2020), or to all reflect laboratory identifier errors, and further analyses are ongoing.

A major study strength is its design, namely being a large-scale community survey recruiting randomly selected private residential households, and testing participants regardless of symptoms. However, its size and scale is also a limitation, since we were not able to collect additional data to comprehensively characterise individual positives. We may have underestimated the initial prevalence of symptoms due to originally asking about current symptoms before July 2020 (subsequently symptoms in the 7 days preceding the visit). As this was only at the earliest visits, mostly weekly, only very transient symptoms between visits would likely have been missed. Similar rates of symptom reporting in the first and last parts of the period analysed suggests that this question was likely generously interpreted in any case. We made no attempt to collect additional information on symptoms after positives were identified to minimise recall bias. This may partly explain why we observed higher rates of positive tests without reported symptoms than recent reviews (Buitrago-Garcia et al., 2020; Byambasuren et al., 2020); however, many studies in these reviews tested close contacts of index cases identified through symptoms and therefore might plausibly have higher viral loads. We compared distributions of Ct values to overall positivity rates in England, since these are the longest series of official statistics available; overall UK positivity estimates are not produced because the four countries making up the UK have different policies and timings regarding community restrictions including lockdowns.

Ultimately, the importance of asymptomatic and low virus-level infections depends on their transmissibility and their prevalence; regardless of limitations in symptom ascertainment, infection without recognition has the potential for onward transmission and unascertained infections are likely critical for avoiding resurgence after lifting lockdown (Hao et al., 2020). Our findings support the use of Ct values and genes detected more broadly in public testing programmes, predominantly testing symptomatic individuals and case contacts, as an ‘early warning’ system for shifts in potential infectious load and hence transmission, and hence the risks posed by individuals to others. This has recently also been proposed on the basis of theoretical work linking effective reproduction numbers to population-level Ct (Hay et al., 2020). In our study, declines in mean and median Ct values preceded or at least coincided with increases in office estimates of positivity rates (Figure 3B); given the far larger numbers that would be available in testing programmes, future research should investigate whether the greater power afforded by continuous outcomes could lead to significantly earlier detection of future positivity increases, particularly within small geographical areas. Ct data are widely available within-laboratory management systems; providing comparisons across the wide variety of commercial assays were interpreted carefully, they could be used alongside available risk factor and symptom information to facilitate more informed and effective individual-level and public health responses to the SARS-CoV-2 pandemic.

Acknowledgements

Office for National Statistics: Iain Bell, Ian Diamond, Alex Lambert, Pete Benton, Emma Rourke, Stacey Hawkes, Sarah Henry, James Scruton, Peter Stokes, Tina Thomas. Office for National Statistics, Analysis John Allen, Russell Black, Heather Bovill, David Braunholtz, Dominic Brown, Sarah Collyer, Megan Crees, Colin Daglish, Byron Davies, Hannah Donnarumma, Julia Douglas-Mann, Antonio Felton, Hannah Finselbach, Eleanor Fordham, Alberta Ipser, Joe Jenkins, Joel Jones, Katherine Kent, Geeta Kerai, Lina Lloyd, Victoria Masding, Ellie Osborn, Alpi Patel, Elizabeth Pereira, Tristan Pett, Melissa Randall, Donna Reeve, Palvi Shah, Ruth Snook, Ruth Studley, Esther Sutherland, Eliza Swinn, Heledd Thomas, Anna Tudor, Joshua Weston. Office for National Statistics, Secure Research Service Shayla Leib, James Tierney, Gabor Farkas, Raf Cobb, Folkert van Galen, Lewis Compton, James Irving, John Clarke, Rachel Mullis, Lorraine Ireland, Diana Airimitoaie, Charlotte Nash, Danielle Cox, Sarah Fisher, Zoe Moore, James McLean, Matt Kerby. University of Oxford, Nuffield Department of Medicine: Ann Sarah Walker, Derrick Crook, Philippa C Matthews, Tim Peto, Emma Pritchard, Nicole Stoesser, Karina-Doris Vihta, Jia Wei, Alison Howarth, George Doherty, James Kavanagh, Kevin K Chau, Stephanie B Hatch, Daniel Ebner, Lucas Martins Ferreira, Thomas Christott, Brian D Marsden, Wanwisa Dejnirattisai, Juthathip Mongkolsapaya, Sarah Cameron, Phoebe Tamblin-Hopper, Magda Wolna, Rachael Brown, Sarah Hoosdally, Richard Cornall, David I Stuart, Gavin Screaton. University of Oxford, Nuffield Department of Population Health: Koen Pouwels. University of Oxford, Big Data Institute: David W Eyre, Katrina Lythgoe, David Bonsall, Tanya Golbchik, Helen Fryer. University of Oxford, Radcliffe Department of Medicine: John Bell. Oxford University Hospitals NHS Foundation Trust: Stuart Cox, Kevin Paddon, Tim James. University of Manchester: Thomas House. Public Health England: John Newton, Julie Robotham, Paul Birrell. IQVIA: Helena Jordan, Tim Sheppard, Graham Athey, Dan Moody, Leigh Curry, Pamela Brereton. National Biocentre Ian Jarvis, Kirsty Howell, Bobby Mallick, Phil Eeles. Glasgow Lighthouse Laboratory Jodie Hay, Harper Vansteenhouse. Department of Health: Jessica Lee. This study is funded by the Department of Health and Social Care. ASW, EP, JVR, TEAP, NS, DE, KBP are supported by the National Institute for Health Research Health Protection Research Unit (NIHR HPRU) in Healthcare Associated Infections and Antimicrobial Resistance at the University of Oxford in partnership with Public Health England (PHE) (NIHR200915). ASW and TEAP are also supported by the NIHR Oxford Biomedical Research Centre. EP and KBP are also supported by the Huo Family Foundation. ASW is also supported by core support from the Medical Research Council UK to the MRC Clinical Trials Unit [MC_UU_12023/22] and is an NIHR Senior Investigator. PCM is funded by Wellcome (intermediate fellowship, grant ref 110110/Z/15/Z) and holds an NIHR BRC Senior Fellowship award. The views expressed are those of the authors and not necessarily those of the National Health Service, NIHR, Department of Health, or PHE. The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

Appendix 1

Appendix 1—figure 1. Relationship between Ct values and viral load.

Appendix 1—figure 1.

Appendix 1—figure 2. Directed acyclic graph of potential relationships between factors.

Appendix 1—figure 2.

*May also depend on factors which effect self-swabbing efficiency, e.g., demographics.

Funding Statement

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

Contributor Information

A Sarah Walker, Email: sarah.walker@ndm.ox.ac.uk.

M Dawn Teare, Newcastle University, United Kingdom.

Miles P Davenport, University of New South Wales, Australia.

Funding Information

This paper was supported by the following grants:

  • Department of Health & Social Care - to A Sarah Walker, Emma Pritchard, Thomas House, Iain Bell, Ian Diamond, Ruth Studley, Jodie Hay, Karina-Doris Vihta, Koen B Pouwels.

  • National Institutes of Health NIHR200915 to A Sarah Walker, Emma Pritchard, Julie V Robotham, Karina-Doris Vihta, Timothy EA Peto, Nicole Stoesser, David W Eyre, Koen B Pouwels.

  • Huo Family Foundation to Emma Pritchard, Koen B Pouwels.

  • Medical Research Council MC_UU_12023/22 to A Sarah Walker.

  • Wellcome Trust 110110/Z/15/Z to Philippa C Matthews.

Additional information

Competing interests

No competing interests declared.

declares lecture fees from Gilead, outside the submitted work.

Author contributions

Conceptualization, Data curation, Formal analysis, Supervision, Methodology, Writing - original draft, Project administration, Writing - review and editing.

Formal analysis, Writing - review and editing.

Conceptualization, Methodology, Writing - review and editing.

Conceptualization, Writing - review and editing.

Conceptualization, Writing - review and editing.

Conceptualization, Supervision, Funding acquisition, Project administration, Writing - review and editing.

Conceptualization, Funding acquisition, Writing - review and editing.

Conceptualization, Funding acquisition, Writing - review and editing.

Conceptualization, Funding acquisition, Project administration, Writing - review and editing.

Funding acquisition, Project administration, Writing - review and editing.

Conceptualization, Methodology, Project administration, Writing - review and editing.

Formal analysis, Writing - review and editing.

Formal analysis, Writing - review and editing.

Conceptualization, Writing - review and editing.

Conceptualization, Formal analysis, Writing - review and editing.

Conceptualization, Writing - review and editing.

Conceptualization, Writing - review and editing.

Conceptualization, Methodology, Writing - review and editing.

Resources; Investigation.

Ethics

Human subjects: Written informed consent was obtained from participants aged 16 years and older, and from parents/carers for those aged 2-15 years; those aged 10-15 years provided written assent. The study received ethical approval from the South Central Berkshire B Research Ethics Committee (20/SC/0195).

Additional files

Supplementary file 1. Month of recruitment into the COVID-19 Infection Survey.
elife-64683-supp1.docx (24.3KB, docx)
Supplementary file 2. Association between characteristics and Ct values.

(A) Univariable effects and main model considering all factors, (B) multivariable model excluding potential mediators of effects of demographics.

elife-64683-supp2.docx (42.2KB, docx)
Supplementary file 3. Data points underlying figures.
elife-64683-supp3.xlsx (47.2KB, xlsx)
Supplementary file 4. Stata code.
elife-64683-supp4.docx (26.3KB, docx)
Transparent reporting form

Data availability

De-identified study data are available for access by accredited researchers in the ONS Secure Research Service (SRS) for accredited research purposes under part 5, chapter 5 of the Digital Economy Act 2017. Individuals can apply to be an accredited researcher using the short form on https://researchaccreditationservice.ons.gov.uk/ons/ONS_registration.ofml. Accreditation requires completion of a short free course on accessing the SRS. To request access to data in the SRS, researchers must submit a research project application for accreditation in the Research Accreditation Service (RAS). Research project applications are considered by the project team and the Research Accreditation Panel (RAP) established by the UK Statistics Authority. Project application example guidance and an exemplar of a research project application are available. A complete record of accredited researchers and their projects is published on the UK Statistics Authority website to ensure transparency of access to research data. For further information about accreditation, contact https://researchaccreditationservice.ons.gov.uk/ons/ONS_homepage.ofml or visit the SRS website. Data points underlying Figures are provided in Supplementary File 4 and Stata code in Supplementary File 3.

References

  1. Adams ER, Ainsworth M, Anand R. Antibody testing for COVID-19: a report from the national COVID scientific advisory panel [version 1; peer review: 1 approved] Wellcome Open Research. 2020;5:139. doi: 10.12688/wellcomeopenres.15927.1. [DOI] [PMC free article] [PubMed] [Google Scholar]
  2. Arkhipova-Jenkins I, Helfand M, Armstrong C, Gean E, Anderson J, Paynter RA, Mackey K. Antibody response after SARS-CoV-2 infection and implications for immunity : a rapid living review. Annals of Internal Medicine. 2021;174:811–821. doi: 10.7326/M20-7547. [DOI] [PMC free article] [PubMed] [Google Scholar]
  3. Buitrago-Garcia D, Egli-Gany D, Counotte MJ, Hossmann S, Imeri H, Ipekci AM, Salanti G, Low N. Occurrence and transmission potential of asymptomatic and presymptomatic SARS-CoV-2 infections: a living systematic review and meta-analysis. PLOS Medicine. 2020;17:e1003346. doi: 10.1371/journal.pmed.1003346. [DOI] [PMC free article] [PubMed] [Google Scholar]
  4. Byambasuren O, Cardona M, Bell K, Clark J, McLaws M-L, Glasziou P. Estimating the extent of asymptomatic COVID-19 and its potential for community transmission: systematic review and meta-analysis. Journal of the Association of Medical Microbiology and Infectious Disease Canada. 2020;5:223–234. doi: 10.3138/jammi-2020-0030. [DOI] [PMC free article] [PubMed] [Google Scholar]
  5. Edwards T, Santos VS, Wilson AL, Cubas-Atienzar A. Variation of SARS-CoV-2 viral loads by sample type, disease severity and time: a systematic review. medRxiv. 2020 doi: 10.1101/2020.09.16.20195982. [DOI]
  6. Endo A, Abbott S, Kucharski AJ, Funk S. Estimating the overdispersion in COVID-19 transmission using outbreak sizes outside China. Wellcome Open Research. 2020;5:67. doi: 10.12688/wellcomeopenres.15842.3. [DOI] [PMC free article] [PubMed] [Google Scholar]
  7. Flaxman S, Mishra S, Gandy A, Unwin HJT, Mellan TA, Coupland H, Whittaker C, Zhu H, Berah T, Eaton JW, Monod M, Ghani AC, Donnelly CA, Riley S, Vollmer MAC, Ferguson NM, Okell LC, Bhatt S, Imperial College COVID-19 Response Team Estimating the effects of non-pharmaceutical interventions on COVID-19 in Europe. Nature. 2020;584:257–261. doi: 10.1038/s41586-020-2405-7. [DOI] [PubMed] [Google Scholar]
  8. Gandhi M, Beyrer C, Goosby E. Masks do more than protect others during COVID-19: reducing the inoculum of SARS-CoV-2 to protect the wearer. Journal of General Internal Medicine. 2020;35:3063–3066. doi: 10.1007/s11606-020-06067-8. [DOI] [PMC free article] [PubMed] [Google Scholar]
  9. Gandhi M, Rutherford GW. Facial masking for Covid-19 — Potential for “Variolation” as We Await a Vaccine. New England Journal of Medicine. 2020;383:e101. doi: 10.1056/NEJMp2026913. [DOI] [PMC free article] [PubMed] [Google Scholar]
  10. Goldman JD, Wang K, Roltgen K, Nielsen SCA, Roach JC, Naccache SN. Reinfection with SARS-CoV-2 and failure of humoral immunity: a case report. medRxiv. 2020 doi: 10.1101/2020.09.22.20192443. [DOI] [PMC free article] [PubMed]
  11. Hao X, Cheng S, Wu D, Wu T, Lin X, Wang C. Reconstruction of the full transmission dynamics of COVID-19 in Wuhan. Nature. 2020;584:420–424. doi: 10.1038/s41586-020-2554-8. [DOI] [PubMed] [Google Scholar]
  12. Hay JA, Kennedy-Shaffer L, Kanjilal S, Lipsitch M, Mina MJ. Estimating epidemiologic dynamics from single crosssectional viral load distributions. medRxiv. 2020 doi: 10.1101/2020.10.08.20204222. [DOI] [PMC free article] [PubMed]
  13. Jacot D, Greub G, Jaton K, Opota O. Viral load of SARS-CoV-2 across patients and compared to other respiratory viruses. Microbes and Infection. 2020;22:617–621. doi: 10.1016/j.micinf.2020.08.004. [DOI] [PMC free article] [PubMed] [Google Scholar]
  14. Jefferson T, Spencer EA, Brassey J, Heneghan C. Viral cultures for COVID-19 infectious potential assessment - a systematic review. Clinical Infectious Diseases. 2020;1:ciaa1764. doi: 10.1093/cid/ciaa1764. [DOI] [PMC free article] [PubMed] [Google Scholar]
  15. Jones TC, Mühlemann B, Veith T, Biele G, Zuchowski M, Hoffmann J, Drosten C. An analysis of SARS-CoV-2 viral load by patient age. medRxiv. 2020 doi: 10.1101/2020.06.08.20125484. [DOI]
  16. Ladhani S, Baawuah F, Beckmann J, Okike I, Ahmad S, Garstang J, Linley E. Prospective Active National Surveillance of Preschools and Primary Schools for SARS-CoV-2 Infection and Transmission in England, June 2020 (SKIDs COVID-19 Surveillance in School KIDs) Phase 1 Report. Public Health England; 2020. [Google Scholar]
  17. Lavezzo E, Franchin E, Ciavarella C, Cuomo-Dannenburg G, Barzon L, Del Vecchio C, Rossi L, Manganelli R, Loregian A, Navarin N, Abate D, Sciro M, Merigliano S, De Canale E, Vanuzzo MC, Besutti V, Saluzzo F, Onelia F, Pacenti M, Parisi SG, Carretta G, Donato D, Flor L, Cocchio S, Masi G, Sperduti A, Cattarino L, Salvador R, Nicoletti M, Caldart F, Castelli G, Nieddu E, Labella B, Fava L, Drigo M, Gaythorpe KAM, Brazzale AR, Toppo S, Trevisan M, Baldo V, Donnelly CA, Ferguson NM, Dorigatti I, Crisanti A, Imperial College COVID-19 Response Team Suppression of a SARS-CoV-2 outbreak in the Italian municipality of Vo'. Nature. 2020;584:425–429. doi: 10.1038/s41586-020-2488-1. [DOI] [PubMed] [Google Scholar]
  18. Lee S, Kim T, Lee E, Lee C, Kim H, Rhee H, Park SY, Son H-J, Yu S, Park JW, Choo EJ, Park S, Loeb M, Kim TH. Clinical course and molecular viral shedding among asymptomatic and symptomatic patients with SARS-CoV-2 infection in a community treatment center in the republic of korea. JAMA Internal Medicine. 2020;180:1447–1452. doi: 10.1001/jamainternmed.2020.3862. [DOI] [PMC free article] [PubMed] [Google Scholar]
  19. Lee LYW, Rozmanowski S, Pang M, Charlett A, Anderson C, Barnard M, Peto TEA. An observational study of SARS-CoV-2 infectivity by viral load and demographic factors and the utility lateral flow devices to prevent transmission. JAMA. 2021;1:ciab421. doi: 10.1093/cid/ciab421. [DOI] [PMC free article] [PubMed] [Google Scholar]
  20. Lumley SF, O'Donnell D, Stoesser NE, Matthews PC, Howarth A, Hatch SB, Marsden BD, Cox S, James T, Warren F, Peck LJ, Ritter TG, de Toledo Z, Warren L, Axten D, Cornall RJ, Jones EY, Stuart DI, Screaton G, Ebner D, Hoosdally S, Chand M, Crook DW, O'Donnell AM, Conlon CP, Pouwels KB, Walker AS, Peto TEA, Hopkins S, Walker TM, Jeffery K, Eyre DW, Oxford University Hospitals Staff Testing Group Antibody Status and Incidence of SARS-CoV-2 Infection in Health Care Workers. The New England journal of medicine. 2021;384:533–540. doi: 10.1056/NEJMoa2034545. [DOI] [PMC free article] [PubMed] [Google Scholar]
  21. Moraz M, Jacot D, Papadimitriou-Olivgeris M. Clinical importance of reporting SARS-CoV-2 viral loads across the different stages of the COVID-19 pandemic. medRxiv. 2020 doi: 10.1101/2020.07.10.20149773. [DOI] [PMC free article] [PubMed]
  22. National SARS-CoV-2 Serology Assay Evaluation Group Performance characteristics of five immunoassays for SARS-CoV-2: a head-to-head benchmark comparison. The Lancet. Infectious Diseases. 2020;20:4. doi: 10.1016/S1473-3099(20)30634-4. [DOI] [PMC free article] [PubMed] [Google Scholar]
  23. Office for National Statistics Coronavirus (COVID-19) Infection survey, UK: 26 march. 2021. [March 26, 2021]. https://www.ons.gov.uk/peoplepopulationandcommunity/healthandsocialcare/conditionsanddiseases/bulletins/coronaviruscovid19infectionsurveypilot
  24. Pouwels KB, House T, Pritchard E, Robotham JV, Birrell PJ, Gelman A, Vihta KD, Bowers N, Boreham I, Thomas H, Lewis J, Bell I, Bell JI, Newton JN, Farrar J, Diamond I, Benton P, Walker AS, COVID-19 Infection Survey Team Community prevalence of SARS-CoV-2 in England from April to November, 2020: results from the ONS coronavirus infection survey. The Lancet Public Health. 2021;6:e30–e38. doi: 10.1016/S2468-2667(20)30282-6. [DOI] [PMC free article] [PubMed] [Google Scholar]
  25. Public Health England Investigation of novel SARS-CoV-2 variant: variant of concern 202012/01; Technical briefing 2. 2020. [July 12, 2021]. https://assets.publishing.service.gov.uk/government/uploads/system/uploads/attachment_data/file/948152/Technical_Briefing_VOC202012-2_Briefing_2_FINAL.pdf
  26. Rasmussen AL, Popescu SV. SARS-CoV-2 transmission without symptoms. Science. 2021;371:1206–1207. doi: 10.1126/science.abf9569. [DOI] [PubMed] [Google Scholar]
  27. Riley S, Ainslie KEC. Transient dynamics of SARS-CoV-2 as England exited national lockdown. medRxiv. 2020a doi: 10.1101/2020.08.05.20169078. [DOI]
  28. Riley S, Ainslie KEC. Resurgence of SARS-CoV-2 in England: detection by community antigen surveillance. medRxiv. 2020b doi: 10.1101/2020.09.11.20192492. [DOI] [PMC free article] [PubMed]
  29. Singanayagam A, Patel M, Charlett A, Lopez Bernal J, Saliba V, Ellis J, Ladhani S, Zambon M, Gopal R. Duration of infectiousness and correlation with RT-PCR cycle threshold values in cases of COVID-19, England, January to may 2020. Euro Surveillance : Bulletin Europeen Sur Les Maladies Transmissibles = European Communicable Disease Bulletin. 2020;25:2001483. doi: 10.2807/1560-7917.ES.2020.25.32.2001483. [DOI] [PMC free article] [PubMed] [Google Scholar]
  30. To KK, Hung IF, Ip JD, Chu AW, Chan WM, Tam AR, Fong CH, Yuan S, Tsoi HW, Ng AC, Lee LL, Wan P, Tso E, To WK, Tsang D, Chan KH, Huang JD, Kok KH, Cheng VC, Yuen KY. COVID-19 re-infection by a phylogenetically distinct SARS-coronavirus-2 strain confirmed by whole genome sequencing. Clinical Infectious Diseases : An Official Publication of the Infectious Diseases Society of America. 2020;1:ciaa1275. doi: 10.1093/cid/ciaa1275. [DOI] [PMC free article] [PubMed] [Google Scholar]
  31. Tomassini S, Kotecha D, Bird PW, Folwell A, Biju S, Tang JW. Setting the criteria for SARS-CoV-2 reinfection – six possible cases. Journal of Infection. 2021;82:282–327. doi: 10.1016/j.jinf.2020.08.011. [DOI] [PMC free article] [PubMed] [Google Scholar]
  32. Walker AS, Vihta K, Gethings O, Pritchard E, Jones J, House T, team C-I. Increased infections, but not viral burden, with a new SARS-CoV-2 variant. medRxiv. 2021 doi: 10.1101/2021.01.13.21249721. [DOI]
  33. Yokota I, Shane PY, Okada K, Unoki Y, Yang Y, Inao T, Sakamaki K, Iwasaki S, Hayasaka K, Sugita J, Nishida M, Fujisawa S, Teshima T. Mass screening of asymptomatic persons for SARS-CoV-2 using saliva. Clinical Infectious Diseases : An Official Publication of the Infectious Diseases Society of America. 2020;1:ciaa1388. doi: 10.1093/cid/ciaa1388. [DOI] [PMC free article] [PubMed] [Google Scholar]

Decision letter

Editor: M Dawn Teare1

Our editorial process produces two outputs: i) public reviews designed to be posted alongside the preprint for the benefit of readers; ii) feedback on the manuscript for the authors, including requests for revisions, shown below. We also include an acceptance summary that explains what the editors found interesting or important about the work.

Acceptance summary:

This paper analyses viral levels (detected as CT) in community acquired SARS-CoV-2 swabs. It suggests a relationship with symptoms, as well as suggesting that declines in observed CT might be used as an indicator in increased transmission.

Decision letter after peer review:

Thank you for submitting your article "Viral load in community SARS-CoV-2 cases varies widely and temporally" for consideration by eLife. Your article has been reviewed by 2 peer reviewers, and the evaluation has been overseen by a Reviewing Editor and Miles Davenport as the Senior Editor. The reviewers have opted to remain anonymous.

The reviewers have discussed their reviews with one another, and the Reviewing Editor has drafted this to help you prepare a revised submission.

Please modify your manuscript to address the comments and recommendations to the authors provided by reviewer 1 and reviewer 2.

Reviewer #1 (Recommendations for the authors):

Overall, the paper contains interesting data, with an exciting statement that increasing viral load can be used as an epidemiological early-warning indicator. They also included a nice attempt of a correlation of Ct with clinical evidence to rank the confidence of positive results. The authors also highlight that some of these RT-PCR-positive cases do not appear to seroconvert and reported possible re-infections despite the presence of anti-spike antibodies which given the current situation with SARS-CoV-2 VOC 202012/01 could be of concern.

However, the interpretation of the fluctuating Ct and their assumption that it can be used as an epidemiological early-warning indicator should be better discuss (see major comments) and more importantly demonstrated.

Furthermore, data on Cts are not completely new as some already described high population-level variability of viral loads without strong correlation with disease severity. However, the data presented here are statistically relevant and therefore interesting. Fluctuating Ct were also previously reported to follow epidemic waves and therefore confinement measures. This should be further discussed here. Finally, some formatting work would be required to better present the data especially regarding the serology results which are neglected in the introduction (see comments).

– In the abstract, the author stated: "Cts changed over time, with declining Ct preceding increasing positivity". This is not clearly shown in the paper. It would be interesting to support this claim to present both Ct and positivity on the same graph to demonstrate that indeed, declining Ct can be used as an early marker of an epidemic wave. From the different graphs presented in Figure 1 this is far from evident. As this observation is the main impact statement, the authors should improve their demonstration as this is critically missing. Percentage of positive test data should not only include the ones obtained in the present study but should be compared with "national data" as your study includes a bias in patients’ selection that might not reflect the "true" situation at the time. Only with this comparison, you can claim that your study design could predict epidemic waves and support your impact statement.

– During the ascending phase of a COVID-19 epidemic wave, patients screened harbor mostly high viral loads, reflecting the start of the disease. Indeed, viral load kinetic for SARS-CoV-2 follow simplistically a sharp increase followed by a relatively slow decline over time. Therefore, with confinement measures in place and lower transmission, it is more likely to screen people at multiple stages of the disease and with therefore lower viral loads. The idea to use mean viral loads as early warning of a start of a new wave is an interesting observation and it would be interesting to place that in the perspective of COVID-19 progression within individuals.

– How long before decrease of Ct can predict and epidemic wave?

– Neither the impact statement nor the abstract mention the serological findings presented in the paper. Is this something deliberate?

– The title does not reflect the content of the paper. If demonstrated, it should at least reflect your impact statement.

– Although described in the material and methods, I would include a short description of the UK's national COVID-19 Infection Survey (CIS) in the introduction to help the reader quickly understand the context of the collected specimens described in this study.

– Fluctuating Ct might likely reflect the confinement measures undertaken by the UK government. In this context, it would be relevant to include the main confinement periods imposed by the government on the different graphs and to discuss their relationship with the viral loads.

– The Results section regarding serological data would require some major improvements, as it is difficult to navigate. It would be much easier for the readers to subdivide the figure in panel to better describe some conclusion. For example, "seven participants appeared to have become infected despite antecedent high anti-spike antibody titres (Figure 4)" should be place in a panel to facilitate the reading (the authors included some color-code but it is still difficult to follow).

– All the figure legends would require some improvement to better guide the reader. For example, supplementary figure 2 is cryptic without a proper legend. And again, I recommend the authors to subdivide their figure in panels, with one panel per "conclusion".

– The chronology of the figures is sometimes cryptic with figures appearing in non-chronological order. I would suggest rethinking the flow of the figures.

– In table 2, it is mentioned that multiple genome were sequenced. Is there any correlations between genome vs symptom, genome vs Cts, ? This would add an interesting part to the paper as these data are likely available.Reviewer #2 (Recommendations for the authors):

Can the authors add some discussion about that they think the different levels of positive are: high, medium and low. I assume that a true positive is someone in whom we find a fragment of RNA that was made by a cell in their body, even if that was some time ago. A false positive is someone for whom there was no real RNA present but through some kind of amplification error or contamination we say that there was? The results seem to suggest that some of the low confidence positives were actual false positives (changing strength of statistical associations) but I didn't feel that the point was fully made.

Some specific points:

Line 95: Couldn't quite understand the sentence on this line

110: What are the units for viral load?

110: Was there any kind of threshold in the definition of positives? Did the threshold differ for single, double of triple positive samples?

115: What is the positivity definition here? Is a true positive someone who has recently produced SARS-CoV-2 RNA and is still testing positive or is it some higher level? Or, is a true negative someone who has not shed SARS-CoV-2 for at least x weeks?

147: How were symptoms defined? Over what period?

151: What were the effects over calendar time?

188: Viral fitness declined over time: I wasn't quite sure what this would be a hypothesis at this point. Maybe needs setting up a little more clearly.

196: Also worth commenting on the cold chain for sample collection, or other study-related factors. A higher ambient temperature for the samples could have led to faster rates of degradation.

222: There could be a wider point here about household transmission. Chickenpox is known to be more severe in household secondaries than in household index, and CT values seem to be correlated with symptoms (https://pubmed.ncbi.nlm.nih.gov/15702036/)

294: Was there a threshold for the single-gene positives?

301: Is there any clear evidence of how the symptom definitions affected reporting when changing from "current" to "previous seven days".

306: As mentioned above, I think some discussion of false positives is merited in this paper. Do the authors feel that detection of an actual RNA fragment from a recent infection is a true or a false positive? Not sure there is an easy answer here, but with all the discussion of detecting infectious people versus just detecting the virus, a bit more context would help.

Page 30 of merged pdf: there is a multipart figure with antibody values on the y axis that I couldn't find a legend for or a figure number.

Supp Figure 2: shouldn't there be an "average true number of virions in the throat and nose" as an unobserved state? This figure might merit a bit more discussion in the intro and Discussion sections.

eLife. 2021 Jul 12;10:e64683. doi: 10.7554/eLife.64683.sa2

Author response


Please modify your manuscript to address the comments and recommendations to the authors provided by reviewer 1 and reviewer 2.

Reviewer #1 (Recommendations for the authors):

Overall, the paper contains interesting data, with an exciting statement that increasing viral load can be used as an epidemiological early-warning indicator. They also included a nice attempt of a correlation of Ct with clinical evidence to rank the confidence of positive results. The authors also highlight that some of these RT-PCR-positive cases do not appear to seroconvert and reported possible re-infections despite the presence of anti-spike antibodies which given the current situation with SARS-CoV-2 VOC 202012/01 could be of concern.

However, the interpretation of the fluctuating Ct and their assumption that it can be used as an epidemiological early-warning indicator should be better discuss (see major comments) and more importantly demonstrated.

Furthermore, data on Cts are not completely new as some already described high population-level variability of viral loads without strong correlation with disease severity. However, the data presented here are statistically relevant and therefore interesting. Fluctuating Ct were also previously reported to follow epidemic waves and therefore confinement measures. This should be further discussed here. Finally, some formatting work would be required to better present the data especially regarding the serology results which are neglected in the introduction (see comments).

– In the abstract, the author stated: "Cts changed over time, with declining Ct preceding increasing positivity". This is not clearly shown in the paper. It would be interesting to support this claim to present both Ct and positivity on the same graph to demonstrate that indeed, declining Ct can be used as an early marker of an epidemic wave. From the different graphs presented in Figure 1 this is far from evident. As this observation is the main impact statement, the authors should improve their demonstration as this is critically missing. Percentage of positive test data should not only include the ones obtained in the present study but should be compared with "national data" as your study includes a bias in patients’ selection that might not reflect the "true" situation at the time. Only with this comparison, you can claim that your study design could predict epidemic waves and support your impact statement.

We have added the requested panel for England (new Figure 3B) where we have the longest follow-up and the largest number of tests, and have added some further text in the Discussion to support this point – which is also clearer with the additional 6 months data included in the revised manuscript. There are no official statistics on positivity rates for the UK as a whole because the 4 countries comprising the UK (England, Wales, Northern Ireland and Scotland) have different policies on community based restrictions enacted at different times, making a combined estimate challenging to interpret – we have added this point as a limitation in the Discussion to explain our focus on England in Figure 3B. We strongly disagree with the reviewer’s assertion that our study includes a bias in patient selection and that there is “national data” that we should compare to. The COVID-19 Infection Survey is the national data, it is the only continuous surveillance study in the UK.

– During the ascending phase of a COVID-19 epidemic wave, patients screened harbor mostly high viral loads, reflecting the start of the disease. Indeed, viral load kinetic for SARS-CoV-2 follow simplistically a sharp increase followed by a relatively slow decline over time. Therefore, with confinement measures in place and lower transmission, it is more likely to screen people at multiple stages of the disease and with therefore lower viral loads. The idea to use mean viral loads as early warning of a start of a new wave is an interesting observation and it would be interesting to place that in the perspective of COVID-19 progression within individuals.

Whilst we agree with the reviewer that this would be interesting, as noted in the limitation section of the Discussion, the large scale and size of the survey means we do not have this kind of specific level of detail on individual disease progression in those testing positive. Further, as we are testing individuals in the community, the numbers that develop sufficiently severe infections to be admitted to hospital or even die is extremely small.

– How long before decrease of Ct can predict and epidemic wave?

In the new Figure 3B we show that after very low rates in the summer of 2020 the decline in Ct values preceded increases in positivity rates in England. However the steep declines in Ct observed in early December 2020 coincided with very sharp increases in positivity, as the B.1.1.7 variant expanded exponentially. We hypothesise that this very fast expansion explains why the increase followed so quickly. After both periods of increases in positivity rates, rising Ct values were accompanied by stabilisation and then declines in the positivity rate.

– Neither the impact statement nor the abstract mention the serological findings presented in the paper. Is this something deliberate?

Yes, this was deliberate given the limit on word count in the abstract (150 words) and the fact that the original manuscript contained preliminary serological data on a small number of participants. However given that there was at least some information available, we felt that we should provide at least limited results. Given the substantially increased numbers, we have amended this short summary, and now report 78% IgG S-antibody positivity after the first RT-PCR-positive and we have amended the abstract to also include this. Our understanding is that the impact statement should be at most 2 sentences and should focus on consequences and therefore we have still not included it there.

– The title does not reflect the content of the paper. If demonstrated, it should at least reflect your impact statement.

We have amended the title to refer additionally to Ct values which we assume is the reviewer’s point. We do not feel that future implications of our findings are appropriate in the title, although would be happy to reconsider if the Editors preferred this.

– Although described in the material and methods, I would include a short description of the UK's national COVID-19 Infection Survey (CIS) in the introduction to help the reader quickly understand the context of the collected specimens described in this study.

We have added this to the last paragraph of the Introduction as suggested.

– Fluctuating Ct might likely reflect the confinement measures undertaken by the UK government. In this context, it would be relevant to include the main confinement periods imposed by the government on the different graphs and to discuss their relationship with the viral loads.

First, we have had to considerably amend this figures to include over 18,000 additional positives that have been detected since the original submission in October. Second, whilst we are presenting a combined analysis of the features of positives across the UK, as described above lockdown measures are determined independently within each nation (England, Wales, Northern Ireland and Scotland) and therefore there is not one specific lockdown measure across the entire population. In response to the reviewer’s comment, and also in the public review, we have added this information to the new panel B in Figure 3 for England only, presenting the official positivity estimates from the survey together with the Ct values for England and with the “stay-at-home” periods clearly marked in different colours according to whether schools were open or shut.

– The Results section regarding serological data would require some major improvements, as it is difficult to navigate. It would be much easier for the readers to subdivide the figure in panel to better describe some conclusion. For example, "seven participants appeared to have become infected despite antecedent high anti-spike antibody titres (Figure 4)" should be place in a panel to facilitate the reading (the authors included some color-code but it is still difficult to follow).

Given the far larger number of positive tests in the study we have completely rewritten this paragraph to provide a much shorter overall summary of the data to date, with a new main Figure 5 summarising the percentage of positive antibody tests over time before and after the first RT-PCR positive swab. Additional more detailed analyses are ongoing and will be presented separately.

– All the figure legends would require some improvement to better guide the reader. For example, supplementary figure 2 is cryptic without a proper legend. And again, I recommend the authors to subdivide their figure in panels, with one panel per "conclusion".

We have added a footnote to Supplementary Figure 2 (now Appendix 1 Figure 2) as suggested, and revisited all the figures, also in the light of the additional data included.

– The chronology of the figures is sometimes cryptic with figures appearing in non-chronological order. I would suggest rethinking the flow of the figures.

We have reordered the figures as suggested; these have also altered somewhat given the substantially larger numbers of positives now included, also spanning the rise and expansion of B.1.1.7.

– In table 2, it is mentioned that multiple genome were sequenced. Is there any correlations between genome vs symptom, genome vs Cts, ? This would add an interesting part to the paper as these data are likely available.

In this analysis, we used whole genome sequencing as an independent confirmation of presence of virus, and have clarified this in the table. Investigating associations between genetic features and outcomes such as symptoms and Ct requires GWAS approaches; as sequences are available for only the minority of positives, and the paper is already relatively lengthy, we consider that this would be best done separately.

Reviewer #2 (Recommendations for the authors):

Can the authors add some discussion about that they think the different levels of positive are: high, medium and low. I assume that a true positive is someone in whom we find a fragment of RNA that was made by a cell in their body, even if that was some time ago. A false positive is someone for whom there was no real RNA present but through some kind of amplification error or contamination we say that there was? The results seem to suggest that some of the low confidence positives were actual false positives (changing strength of statistical associations) but I didn't feel that the point was fully made.

Some specific points:

Line 95: Couldn't quite understand the sentence on this line.

We have reworded to try to improve clarity and removed the abbreviation.

110: What are the units for viral load?

DC indicated direct copies from the QA panel, we have replaced with copies/ml in the main text and defined in Appendix 1 Figure 1 footnote.

110: Was there any kind of threshold in the definition of positives? Did the threshold differ for single, double of triple positive samples?

As described in the Methods, positivity for each gene is determined by the UgenTec Fast Finder 3.300.5 algorithm, which is FDA accredited. We have clarified both here and in the Methods that there is no specific Ct threshold for determining positivity.

115: What is the positivity definition here? Is a true positive someone who has recently produced SARS-CoV-2 RNA and is still testing positive or is it some higher level? Or, is a true negative someone who has not shed SARS-CoV-2 for at least x weeks?

We have clarified here that we are talking about genuine presence of virus, which we would argue is the most relevant measure for a surveillance study. We have also added further Discussion on this important point, in response to the reviewer’s comment below.

147: How were symptoms defined? Over what period?

This information is provided in the Methods; as symptoms are referred to at multiple places in the Results before this, we have added a reference to the Methods the first time this occurs.

151: What were the effects over calendar time?

These are given in the second part of this sentence “with markedly fewer positives with Ct <30 (Figure 1B), very low percentages with symptoms at/around positive tests, and more “lower” evidence positives in July/August” – the point being made is that Ct, symptoms and level of evidence all follow similar patterns over calendar time. (Note this figure is now Figure 3, having reordered as suggested by the first reviewer above.)

188: Viral fitness declined over time: I wasn't quite sure what this would be a hypothesis at this point. Maybe needs setting up a little more clearly.

We have clarified this as suggested – at least in the UK, numerous individuals postulated that declining positivity during July/August 2020 was a consequence of declining viral fitness as the virus adapted to human hosts.

196: Also worth commenting on the cold chain for sample collection, or other study-related factors. A higher ambient temperature for the samples could have led to faster rates of degradation.

We have noted this point here in the Discussion, and also added details on the transport of swabs from household to laboratory to the Methods.

222: There could be a wider point here about household transmission. Chickenpox is known to be more severe in household secondaries than in household index, and CT values seem to be correlated with symptoms (https://pubmed.ncbi.nlm.nih.gov/15702036/)

A separate analysis specifically investigating household transmission is about to be submitted. Given the length of the existing manuscript, we would prefer not to address this further here.

294: Was there a threshold for the single-gene positives?

No, there is no specific threshold for determining positivity (clarified in the Methods). The UgenTec Fast Finder 3.300.5 algorithm incorporates multiple aspects of the individual Ct curves to determine positivity.

301: Is there any clear evidence of how the symptom definitions affected reporting when changing from "current" to "previous seven days".

This information is implicitly presented in Figure 3C (was 1C in the original submission) which shows percentage of positives reporting symptoms over calendar time. Symptom reporting dropped markedly towards the end of June 2020 concurrent with declines in Ct, the questionnaires were changed mid July 2020, and then symptoms rose markedly from mid-August concurrent with increases in Ct. The challenge therefore is that the questionnaire change is confounded with changing Ct over calendar time. As the vast majority of data in the revision is after July 2020 we have not elaborated further on this.

306: As mentioned above, I think some discussion of false positives is merited in this paper. Do the authors feel that detection of an actual RNA fragment from a recent infection is a true or a false positive? Not sure there is an easy answer here, but with all the discussion of detecting infectious people versus just detecting the virus, a bit more context would help.

As suggested, we have expanded our discussion on this in the paragraph originally starting “Since RT-PCR and antigen assays test for viral presence, it is more relevant to consider limits of detection, rather than “false-positives” per se”. We would argue that it is essential to distinguish the difference between a test result and the action arising from the test result, and also the context of the study – for surveillance our goal is not detecting infectious people in the same way that symptomatic testing (with subsequent isolation, contact tracing etc) needs to. Therefore there are different “gold standards”.

Page 30 of merged pdf: there is a multipart figure with antibody values on the y axis that I couldn't find a legend for or a figure number.

All figures are now provided as separate tif files including Figure titles and legends.

Supp Figure 2: shouldn't there be an "average true number of virions in the throat and nose" as an unobserved state? This figure might merit a bit more discussion in the intro and Discussion sections.

We have added this additional unobserved state to Supplementary Figure 2 (now Appendix 1 Figure 2) as suggested, and also added a paragraph to the Discussion referring to this figure and the updated related findings.

Associated Data

    This section collects any data citations, data availability statements, or supplementary materials included in this article.

    Supplementary Materials

    Supplementary file 1. Month of recruitment into the COVID-19 Infection Survey.
    elife-64683-supp1.docx (24.3KB, docx)
    Supplementary file 2. Association between characteristics and Ct values.

    (A) Univariable effects and main model considering all factors, (B) multivariable model excluding potential mediators of effects of demographics.

    elife-64683-supp2.docx (42.2KB, docx)
    Supplementary file 3. Data points underlying figures.
    elife-64683-supp3.xlsx (47.2KB, xlsx)
    Supplementary file 4. Stata code.
    elife-64683-supp4.docx (26.3KB, docx)
    Transparent reporting form

    Data Availability Statement

    De-identified study data are available for access by accredited researchers in the ONS Secure Research Service (SRS) for accredited research purposes under part 5, chapter 5 of the Digital Economy Act 2017. Individuals can apply to be an accredited researcher using the short form on https://researchaccreditationservice.ons.gov.uk/ons/ONS_registration.ofml. Accreditation requires completion of a short free course on accessing the SRS. To request access to data in the SRS, researchers must submit a research project application for accreditation in the Research Accreditation Service (RAS). Research project applications are considered by the project team and the Research Accreditation Panel (RAP) established by the UK Statistics Authority. Project application example guidance and an exemplar of a research project application are available. A complete record of accredited researchers and their projects is published on the UK Statistics Authority website to ensure transparency of access to research data. For further information about accreditation, contact https://researchaccreditationservice.ons.gov.uk/ons/ONS_homepage.ofml or visit the SRS website. Data points underlying Figures are provided in Supplementary File 4 and Stata code in Supplementary File 3.


    Articles from eLife are provided here courtesy of eLife Sciences Publications, Ltd

    RESOURCES